Performance predictions for the Robert Stobie Spectrograph near infrared arm on SALT

Performance predictions for the Robert Stobie Spectrograph near infrared arm on SALT Marsha J. Wolf a*, Andrew I. Sheinis a, Theodore B. Williams b, Kenneth H. Nordsieck a, Matthew A. Bershady a a University of Wisconsin, Astronomy Dept., 475 N. Charter St., Madison, WI 53706 b Rutgers University, Dept. of Physics & Astronomy, 136 Frelinghuysen Rd., Piscataway, NJ 08854 ABSTRACT The Robert Stobie Spectrograph near infrared arm will provide high throughput, low to medium resolution long slit and multi-object spectroscopy with broadband, spectropolarimetric, and Fabry-Perot imaging modes over a 8 diameter field of view. The wavelength range of the instrument is 0.9-1.7 microns, and can be operated simultaneously with the visible arm to extend the short wavelength limit to 0.32 microns. Once fielded, RSS-NIR will be the only facility instrument on an 8-10 meter class telescope with multi-object spectroscopy capability covering this spectral range simultaneously. RSS-NIR is scheduled to be commissioned on the 11-meter Southern African Large Telescope in late 2012. This is an upgrade to the existing visible instrument, with which it shares the slit plane and an ambient temperature collimator. Beyond the collimator, the NIR arm is cooled to -40 o C, with a cryogenic dewar containing the detector, long wavelength blocking filters, and final camera optics. This semi-warm configuration has required extensive upfront analysis of the instrumental thermal background levels, which have been incorporated into the instrument performance simulator. We present the performance predictions for spectroscopic modes of RSS-NIR and preliminary performance estimates and NIR issues still being addressed in the design for Fabry-Perot and polarimetric modes. Keywords: spectrograph, near infrared, Fabry-Perot, spectropolarimetry, SALT, thermal background, performance simulations 1. INTRODUCTION From the beginning, the visible Robert Stobie Spectrograph (RSS-VIS) 1 on the Southern African Large Telescope (SALT) 2 has envisioned an additional near infrared (NIR) arm. This extension of the instrument, RSS-NIR, 3 will be the first NIR instrument on SALT, expanding the capabilities of the telescope into an entirely new regime. With the exception of X-shooter on the VLT, the combined RSS will be unique among instrumentation for 8-10 meter class telescopes in its ability to simultaneously record data in the UV-visible and NIR. The RSS-NIR upgrade will include a number of operational modes over the λ = 0.9 to 1.7 µm wavelength range: imaging, very high throughput low to medium resolution spectroscopy, and narrowband Fabry-Perot imaging. In addition, polarimetry can be added to any of these modes. Polarimetric modes are limited to a field of view of 4 x 8 arcmin, while all other modes utilize the full 8 arcmin diameter field. The design incorporates an articulated camera, volume phase holographic (VPH) gratings, and a single etalon Fabry-Perot system. An instrument layout is shown in Figure 1. This instrument leverages the considerable effort and expense undertaken by UW researchers and others for the visible system, while preserving all of the visible capability. The design philosophy for RSS-NIR was to duplicate the capabilities of the visible side where possible, with any necessary adaptations for operation in the NIR. The operational modes of RSS-NIR fall into 5 major categories: 1) imaging, 2) long slit spectroscopy, 3) multi-object spectroscopy, 4) Fabry-Perot tunable narrow band imaging, and 5) polarimetry, either imaging or spectroscopic. Each of these modes uses a variety of mechanisms in the instrument that control insertion of: 1) the focal plane long slits and multi-object masks; 2) the dispersive elements, consisting of either gratings or the Fabry-Perot etalon; 3) the polarizing * mwolf@astro.wisc.edu

optics, two waveplates and a polarizing beamsplitter, and 4) filters consisting of broadband and narrowband filters in the pre-dewar (-40 o C) and long wavelength cutoff filters in the cryogenic dewar (120 K). Permutations of these mechanisms combine to give RSS-NIR the versatility to address a broad range of astrophysical research programs to be observed on the queue-scheduled SALT telescope. Figure 1. RSS instrument layout. The common components between the visible and NIR arms reside in an ambient temperature zone and include the slit, insertable waveplates for polarimetric observations, the collimator, and the dichroic beamsplitter. The visible beam is reflected by the dichroic and the components of that arm are shown as in gray on the right. The NIR beam is transmitted through the dichroic to its final doublet element of the collimator. The NIR doublet serves as the window into a pre-dewar enclosure that operates at -40 o C and is purged with slightly overpressure dry air. Components within the pre-dewar include a fold mirror, order sorting filters for Fabry-Perot mode, the Fabry-Perot etalon, the spectroscopic gratings, the polarizing beamsplitter, the first 5 camera optics, and mounts and insertion actuators for all of these components. A cryogenic vacuum dewar is enclosed within the pre-dewar. It houses the HAWAII-2RG detector with SIDECAR ASIC electronics, long wavelength blocking filters, and the last 2 camera optics, one of which serves as its entrance window. Both instruments sit on the existing space frame at the prime focus of SALT that was constructed with RSS-VIS. RSS-NIR is in the final design phase, with its critical design review scheduled for October 2010, and instrument delivery to occur in late 2012. 2. THERMAL STRAY LIGHT ANALYSIS One of the largest factors in maximizing the performance of an instrument such as this is minimizing the instrumental thermal background. RSS-NIR operates in a partially cooled regime between fully cryogenic and ambient temperature, where a number of spectrographs have not been entirely successful due to high thermal backgrounds. Therefore, detailed predictive thermal stray light analysis is a high priority for this project and is integral to the entire design. Before discussing results of the thermal stray light analysis, a few issues related to the ambient conditions at the telescope site must be considered and are covered in the next two sections.

2.1 Environmental Conditions at SALT Because part of RSS-NIR operates at ambient temperature, it is important to consider typical operating conditions at the site. Environmental data collected in 2007 and 2008 are shown in Figure 2. The hourly median and maximum summer (Dec-Feb) temperatures are 14 and 23 o C, while the median and minimum winter (Jun-Aug) temperatures are 3 and -4 o C. These data are used to determine the range of temperatures over which ambient components must be analyzed, and the amount of time during the year that we can expect these conditions to occur. Figure 2. Top Left: Monthly night time average temperatures at the SALT site, as measured from a 10-meter high external tower over the period of January 2007 to February 2008. Top Right: The percentage of hours between 18 o twilight that fall into different temperature bins over the same time period. Bottom: Relationship between temperature, relative humidity, and due point at the telescope. 2.2 The NIR Night Sky Our goal for RSS-NIR spectroscopy is that the instrument thermal background be below the night sky continuum level in between OH emission lines. However, no near infrared spectra of the night sky at SALT exist. Therefore, we have been using the observed sky spectrum from Maunea Kea in Hawaii at the Gemini Observatory (http://www.gemini.edu/?q=node/10781) shown in Figure 3. The SALT telescope is located in Sutherland, South Africa, at 20 o 48 38.5 E, 32 o 22 46 S, at an elevation of 1798 m. We are investigating the possibility of using atmospheric models to predict the sky at SALT. Attempts to reproduce the observed spectrum at Maunea Kea will check the feasibility of using the models to generate an appropriate sky spectrum for SALT using radiosonde data collected around South Africa as inputs to the models. We also plan to obtain some red sky spectra at R ~ 8000 during RSS-VIS recommissioning in late 2010. For now, we use the Maunea Kea sky spectrum. We smooth it to spectral resolutions covering the range of RSS-NIR and measure the continuum level between OH emission lines for the different resolutions at the specific wavelengths to be

analyzed. These sky background levels (see Figure 7) are used to determine the values below which our instrument thermal background must lie in order to allow sky-limited observations of faint astronomical objects. Figure 3. The night sky spectrum from Maunea Kea at the Gemini Observatory. Observing conditions were airmass = 1.5 and H 2 O = 1.6 mm with spectral resolution of λ = 0.04 nm (R ~ 33750). Broadband NIR sky measurements for both Maunea Kea and Sutherland are marked with the symbols denoted in the legend. 2.3 Stray Light Analysis Our analysis is performed using the Advanced Systems Analysis Program (ASAP) by Breault Research Organization. The optical design of SALT, RSS, and the NIR arm are imported into ASAP from Zemax files and the mechanical designs of mounts and structures are imported from SolidWorks CAD files. Any component can be made into a thermal emitter with the proper temperature, emissivity, and scattering characteristics. We then perform Monte Carlo nonsequential ray traces of this thermal radiation through the system. Initially, our model of the NIR arm in ASAP was largely conceptual, but allowed us to roughly determine required operating temperatures of regions of components early in the project. 4 With this early model we set the nominal pre-dewar operating temperature at -40 o C. As the mechanical design and thermal FEA modeling of the instrument matures, the ASAP model is being used at a more detailed level to design baffles, the cold pupil mask, and radiation shields within the cooled areas. Preliminary results of the instrument thermal backgrounds predicted from ASAP have been incorporated into the instrument performance simulations presented in this paper. The first step in such an analysis is to determine which surfaces in the telescope and instrument systems emit thermal radiation that reaches the detector. This is done by placing a source at the detector and tracing rays backwards through the system to determine a list of critical objects, scaled in relative importance by their geometrical configuration factors (the view factors between each object and the detector). Additionally, every inner surface of lens mounts from which radiation could couple into a lens, is set up as a thermal emitter. Collimator mount surfaces are shown in Figure 4. The highest contributor to thermal stray light in spectroscopy mode is the slit because it sits directly in the beam. The long slits, shown in Figure 5 consist of a stainless steel slit blank that is assembled in its mount at an 11 angle so that a slit viewing camera has access to the bottom side. The blank has a 1 arcsec wide slit cut into it, which is typical for the SALT site seeing. Baffles cover the field of view on either side of the slit blank. Because of the angled slit, stray thermal

radiation can take the form of reflections off of specific areas that indirectly reach the detector. An example is shown in Figure 5. Many such indirect stray light paths would not be obvious without this type of analysis. Figure 4. The yellow highlights indicate inner surfaces of collimator lens mounts in the main group that are set up to be thermal emitters in the model. Figure 5. Example of an indirect thermal stray light path originating as emission from a flat ring on the bottom of a lens mount that reflects off of the long slit at the bottom, which sits at an 11 o angle relative to the optical axis, and propagates directly up the optical train to the detector. The collimator main group begins at the top. The telescope is modeled as a circular 11 meter diameter primary mirror, but with the effective thermal emissivity of the actual segmented mirror. For this estimate we have assumed an average width of 2 cm for the 240 gaps between 91

hexagonal segments. This gives a mirror area of 95 m 2 and a gap area of 2.8 m 2. Our estimate of the effective primary mirror emissivity, scaled by relative areas of segments and gaps, is ε = 0.125. A mirror surface emissivity of ε = 0.1 is assumed for both the primary mirrors and the 4 spherical aberration corrector (SAC) mirrors, to simulate dirty mirrors. SALT is a fixed azimuth telescope. The 4-mirror SAC and instrument package at prime focus is moved by a tracker during an observation, scanning the optical pupil across the primary mirror. Two configurations are shown in Figure 6 and have been incorporated into the ASAP model: one at the center of a track with the telescope pupil centered on the primary mirror and one at the outer extreme of a track with the pupil partially off the primary. In the extreme edge position, the floor is included as a thermal emitter. Figure 6. Two telescope configurations analyzed. Left: tracker and instrument package at the center of the primary mirror. Right: tracker at the outer extreme. In this case the floor (purple) is included as a thermal emitter. The gold wireframe indicates the position of the telescope pupil in this configuration. Other emission sources in the model include the ambient temperature optics, the optical tower around the dicroic beamsplitter, emission from the visible arm coupled through the dichroic beamsplitter, and the pre-dewar enclosure and all components within it operating at -40 o C. Yet to be added are the telescope structure, the dome, and mounts and optical baffles for the SAC mirrors. 2.4 Instrument Thermal Background Results Figure 7 shows the predicted thermal radiation reaching the detector from the ambient temperature components at our longest operational wavelength, λ = 1.7 µm, as a function of temperature. Values drop by a factor of approximately 6 for a long wavelength cutoff of λ = 1.6 µm. The vertical lines mark median and max/min seasonal temperatures at the site. The dashed horizontal lines mark our estimated sky continuum level between OH emission lines for different spectral resolutions, assuming an instrument throughput of 0.3 for the sky emission. The most challenging observational conditions are spectroscopy of faint sky-limited objects at high spectral resolution during warm ambient temperatures. Three different emissivities for the slit were considered, ε = 0.95 for a black slit, ε = 0.06 for a shiny aluminum slit, and ε = 0.02 for a shiny gold slit. It is clear that a low emissivity gold slit provides much better results, but even that would not allow sky-limited operation out to 1.7 µm. A range of emissivities between 1 and 0.1 were considered for the thermal radiation coupled in from the visible beam. The worst case would put the visible beam as the second largest contributor, at a factor of 10 below the gold slit. Next in line are the ambient temperature optics (taken here as a whole) at a factor of 30 below the gold slit. Ambient temperature optic mounts are next in importance, with the optical tower and telescope, even at the edge of a track, negligible. The contribution from the pre-dewar and its contents is of order 4e-15 J s -1. Because the slit is the largest contributor to the thermal background at the detector, we considered cooling the slit. The resulting total instrument thermal background reaching the detector with a cooled slit is shown in Figure 8 for different dt s below ambient.

Figure 7. Instrument thermal background from different ambient temperature components predicted by ASAP. Figure 8. Instrument thermal background at the detector with a gold slit cooled to different dt s below ambient. Using the historical annual site temperatures during night time observing hours (Figure 2), slit cooling would allow increased operation as shown in Figure 9. The left plot is for a long wavelength cutoff of λ = 1.7 µm, and right is for

1.65 µm. With no cooling, operation out to 1.7 µm would never be possible at any spectral resolution. Cooling to 20 o C below ambient would allow operation out to 1.7 µm during about half of the winter, and cooling to 30 o C below ambient would extend this into the summer nights that are cooler than the median summer temperature. For a cutoff wavelength of 1.65 µm, cooling to 20 o C below ambient would allow observations to occur at least 90% of the time. Given these results, we have built up a slit cooling prototype in the lab using a TEC cooler. Cooling to 20 o C below ambient has been demonstrated and further investigations should lead to improvements. 5. The next step is to test the cooling in humidity and due points representative of conditions at the telescope ( Figure 2) to verify that we can mitigate condensation with dry purge air. Figure 9. The percentage of night time observing hours between 18 o twilight for which site conditions would allow sky-limited spectroscopy at different spectral resolving powers, R, to occur with the slit cooled to different levels below ambient temperature. Left: Cutoff wavelength of 1.7 µm. Right: Cutoff wavelength of 1.65 µm. 3. SYSTEM THROUGHPUT We will have a suite of 4 volume phase holographic gratings (VPHGs) to cover the wavelength and spectral resolution ranges of the instrument. Efficiency contours predicted by rigorous coupled wave analysis (RCWA) 6 are shown in Figure 10 (left). The plot on the right shows the predicted throughput of various components of the instrument. The optics curve (blue) includes the existing collimator (6 lenses and a waveplate compensator made of CaF 2, fused quartz, and NaCl) and all RSS-NIR collimator (doublet made of fused quartz and CaF 2 ) and camera optics (7 lenses made of fused quartz, CaF 2, S-LAH60, BaF 2, I-FPL51Y, and S-NPH53). The dichroic beamsplitter curve (cyan) is the preliminary predicted performance by one of our potential vendors, not yet completely optimized as seen by the efficiency falloff beyond 1.5 µm. The detector curve is the Teledyne measured quantum efficiency of our science grade HAWAII-2RG array. The detector cutoff wavelength is 1.73 µm. VPHG efficiencies are given both at a fixed resolving power, R=4000 (solid green), and for the grating superblaze functions (dashed green). Total instrument throughput for spectroscopy is shown by the thick magenta lines (solid for R=4000, dashed for VPHG superblaze). The thick black line includes the telescope, assuming mirror reflectivities of 0.89 for the primary and 0.96 for the SAC. At R = 4000, we predict an overall instrument throughput of 40-60% at the peak of the grating blaze functions and 20-40% at the edge of the blaze functions. 4. SPECTROSCOPY We have incorporated all previous results into an instrument performance simulator for RSS-NIR. It uses the observed Maunea Kea night sky spectrum, the predicted instrumental thermal backgrounds at the detector, and the throughput of all optics to predict on-sky performance. Figure 11 shows predicted limiting Vega magnitudes as a function of wavelength to reach S/N = 10 in a 1 hour integration with 1 arcsec seeing and an ambient temperature of +20 o C. The different symbols represent different levels of slit cooling. A read noise of 6 e - is assumed, which is our expected goal using Fowler or up-the-ramp sampling techniques. The plot on the left is for a spectral resolution of R = 4000 and the right is for R = 7000. With no slit cooling the predicted median limiting magnitudes for R = 4000 (7000) are J ~ 21.4

(20.6) and H short ~ 21.0 (20.4). (H short is defined over the H bandpass out to our longest wavelength cutoff of λ = 1.7 µm.) Slit cooling to dt ambient = -30 o C would improve the limiting H magnitudes by 0.1 at λ = 1.5 µm to 0.6 at λ = 1.7 µm. Figure 10. Left: RCWA predicted efficiency contours (average of s and p polarizations) for our 4 VPHGs. Contours represent efficiencies of 80% (solid), 60% (dashed), and 40% (dotted). Right: Predicted instrument throughput due to various components: optics (blue), dichroic beamsplitter (cyan), detector measured quantum efficiency (red), VPHG efficiency at R=4000 (solid green), and VPHG superblaze efficiency (dashed green). Total instrument throughput for spectroscopy is shown by the thick magenta lines (solid for R=4000, dashed for VPHG superblaze). The thick black line includes estimated telescope efficiency. Figure 11. Predicted instrument performance for spectroscopic observations. These are limiting Vega magnitudes as a function of wavelength to reach S/N = 10 in a 1 hour integration assuming 1 arcsec seeing and T ambient = 20 o C. The red squares are with no slit cooling, the blue triangles are with the slit cooled to dt ambient = -20 o C, and the green circles are with the slit cooled to dt ambient = -30 o C. Left is for R = 4000 and right is for R = 7000. The median values in the upper right corners are for no slit cooling. 5. FABRY-PEROT SPECTRAL IMAGING The Fabry-Perot mode employs a single 150 mm clear aperture etalon, cooled to -40 o C and located in the collimated beam. The transmitted wavelength is selected by adjusting the separation of the etalon plates. The spacing and parallelism of the etalon are measured with capacitive sensors optically attached to the plates, and piezoelectric actuators

controller by a servo system to set and maintain the desired spacing. The preliminary etalon design will have a gap of 33 μm, plate reflectivity of 95%, and surface flatness of λ/150 or better. This will produce a spectral resolution of R = 2500 and a separation between adjacent orders of 24 nm at λ = 1.25 μm. At this resolution we would use order sorting filters of bandwidth 17 nm (at λ = 1.25 μm) to select the desired order of the etalon and reject parasitic light from adjacent orders to less than 2%. The visible system with similar configuration has a throughput of 60%. 7 A few issues regarding the operation of a NIR Fabry-Perot etalon are still being analyzed. First is the feasibility of operating an etalon with piezoelectric actuators and capacitive gap sensors at -40 o C inside our pre-dewar. We are currently thermal cycling a test device from IC Optical Systems in a laboratory environmental chamber to evaluate its endurance characteristics at this temperature. We expect operation at -40 o C to be more successful than for some previous cryogenic temperature devices, however, if any problems arise we will work with the vendor to address them. A second issue is the effect of the typical extended wings of the etalon transmission profile (resembling a Voigt profile) in the presence of numerous strong night sky lines. Because we are using a single etalon system, a departure from the visible side, careful modeling must be done to verify whether sky lines can be sufficiently blocked by the combination of the etalon and an order blocking filter. We know that this will not be possible in every part of the observable spectrum, and plan to select windows in quieter parts of the sky spectrum in which observations will be possible. An example of possible filter locations in the J-band is shown in Figure 12 (this does not yet have the etalon transmission profile applied). We are investigating the severity of the line wing issue with detailed modeling and on-sky observational tests using the visible arm with single and double etalons at its reddest wavelengths, λ 0.9 µm. Figure 12. Possible Fabry-Perot observing windows in the J-band. The black spectrum is the Maunea Kea night sky smoothed to R = 2500 (with a gaussian lineshape, not the etalon transmission profile), the blue curve near the top shows relative atmospheric transmission, the green vertical lines mark spectral features of astronomical interest, the cyan line near the bottom marks a factor of 2 above the sky continuum, the narrow red dashes mark regions with a width of 5 etalon linewidths that contain no night sky lines above the cyan line (clear spectral windows), and the thick magenta horizontal lines mark possible order blocking filter locations. NIR Fabry-Perot emission line observations, assuming a sky background at the continuum level between gaussian profile sky emission lines, would have the sensitivity limits shown in Figure 13. This shows that we could detect sources with modest star formation rate in Lyα to z = 12. Other nebular diagnostics are easily detectable at lower redshift, thus opening star formation and abundance studies to more quiescent galaxies to z = 2.5. Our initial estimates of the increase in sky background due the etalon transmission wings is a maximum of a factor of 5 at λ ~ 1 µm and a factor of 20 at λ ~ 1.65 µm. This would decrease the emission line sensitivity by factors of approximately 2 and 5, respectively, which would limit Lyα studies to z ~ 8 and star formation and abundance studies in more quiescent galaxies to z ~ 1.5-2. We expect that careful selection of blocking filter central wavelengths, spectral bandpasses, and tighter blocking specifications will decrease this sky background significantly in some regions throughout the spectral range. Analyses are ongoing to determine the level of realizable improvement.

Figure 13. Estimated Fabry-Perot sensitivity limits for 5σ emission-line detection in a 1 hr exposure at R = 4000 within 1 arcsec 2 for the RSS VIS and NIR beams (grey areas and dashed curves). This assumes a sky background at the continuum level between gaussian profile sky emission lines. Limits correspond to regions between skylines, ~40-60% of the NIR band-pass in J and H at this resolution. Atmospheric extinction is also shown. Labeled redshift tracks for Lyα through Hα for a galaxy forming 1 M sun yr -1 of stars. 6. SPECTROPOLARIMETRY The performance of RSS-VIS in polarimetric modes has been initially tested. Polarimetric efficiency has been measured at > 95% for linear polarization and > 94% for circular polarization. Transmission has been measured as > 70% of the spectroscopic and imaging RSS modes at λ = 0.65 µm. Instrumental polarization has not yet been completely characterized, but the goal is to achieve < 0.4% for linear and < 3e-3 linear to circular ratio. There are three issues to be resolved in accomplishing the same capabilities for RSS-NIR: (1) the detailed construction of the Wollaston prism polarizing beamsplitter (a mosaiced array of 9 prism pairs), (2) the specification requirements for the dichroic beamsplitter, and (3) the uniformity of polarization efficiency across the clear aperture of the VPH gratings. The Wollaston beamsplitter on the visible arm has lens fluid coupling between the two prism wedges of each mosaic element, both to reduce ghost images at the prism interface and to allow for coalignment of the mosaic by adjusting the fluid wedges individually until the mosaic lines up. The latter was advantageous because the prism manufacturer, Karl Lambrecht Corp, could not guarantee matching the mosaic wedges to within the imaging specification. For the NIR side, a fluid interface is considered to be risky at the -40 C temperature of the pre-dewar, where the beamsplitter is located, so the prism wedges will be air-coupled. The ghosts will be controlled by AR coatings (which are easier in the NIR than in the UV-visible), and Karl Lambrecht has quoted for accomplishing the wedge matching by fabricating enough wedges that selection allows the specification to be met. This is feasible in the NIR since NIR quality calcite is much easier to obtain than UV quality calcite. The issue of the dichroic beamsplitter arises because in the past some dual-beam instruments used for spectropolarimetry (Keck LRIS and Lick Kast) have shown instrumental polarimetric calibration stability problems at the 0.1% level. These problems went away when those instruments were used in single-beam mode. This clearly must be a function of the dichroic polarimetric properties, which can be controlled by specification. The goal for RSS-NIR will be to reduce the uncalibratable systematic instrument polarization to 0.03%, a factor of three better than existing instruments. As a backup, the mechanical design will allow for manual replacement of the dichroic by a mirror or window, to allow for single beam operation in campaign mode for the highest precision spectropolarimetry on either arm. Preliminary observed data from RSS-VIS hint at a contribution to systematic instrumental polarization due to nonuniformities across the clear aperture in the VPHG efficiencies for s and p polarizations at the > 5% level. Our specification for the RSS-NIR VPH gratings will include a tighter specification on this uniformity, as well as required

vendor tests to demonstrate this performance. Vendors have estimated their manufacturing capabilities to be within 3-5% polarization uniformity for our suite of gratings. 7. SUMMARY We have presented instrument performance predictions for RSS-NIR spectroscopy. Overall instrument throughput for this mode is predicted to be 40-60% at the peak of the grating blaze functions and 20-40% at the edge of the blaze functions for R = 4000. On-sky performance predictions incorporate extensive simulations of the expected instrument thermal background. The thermal background is estimated by performing a non-sequential ray trace of thermal radiation emitted from all surfaces in the instrument that can be seen by the detector. To determine when sky-limited astronomical observations would be possible, the instrument background levels at different ambient temperatures are compared to an estimated sky background, which is the night sky continuum level in between OH emission lines taken from an observed spectrum of the night sky at Maunea Kea in Hawaii (since no such data currently exist from the SALT site). Using historic temperature data from the site, the amount of time that sky-limited spectroscopic observations would be possible out to different long wavelength cutoffs is quantified for different levels of slit cooling (the highest contributor to instrument thermal background). With no slit cooling sky-limited spectroscopy would never be possible out to the longest cutoff wavelength of 1.7 µm, but could be done out to λ = 1.65 µm ~20% of the time. Cooling the slit to 30 o C below ambient would allow observations out to λ = 1.7 µm ~50% of the time. On-sky limiting magnitudes to reach S/N = 10 in 1 hour with 1 arcsec seeing at R ~ 4000 (7000) are predicted to be J ~ 21.4 (20.6) and H short ~ 21.0 (20.4). Slit cooling to dt ambient = -30 o C would improve the limiting H magnitudes by 0.1 at λ = 1.5 µm to 0.6 at λ = 1.7 µm. Fabry-Perot performance is expected to be similar to the visible arm, with a throughput of ~60%. We are still analyzing how much the etalon transmission profile wings will affect the NIR sky background in the presence of numerous strong OH emission lines. Initial estimates indicate that the sky continuum level could be up by maximum factors of 5 and 20 at λ = 1 and 1.65 µm. Evaluation of the severity of this effect is continuing through red sky observations with the existing visible Fabry-Perot instrument and more complete modeling of the system. We will tailor the blocking filter central wavelengths, bandpasses, and blocking level specification to maximize performance in the most sky windows. Spectropolarimetry performance is also expected to be similar to the visible arm, which has a polarimetric efficiency of > 94% and transmission > 70% of the imaging and spectroscopic instrument modes. Issues that could potentially affect the performance of this mode in the NIR include the polarization uniformity performance of the dichroic beamsplitter and the VPH gratings. These characteristics are being dealt with by tightening the specifications for these components in conjunction with the expected manufacturing capabilities of the vendors. REFERENCES [1] Burgh, Eric B., Nordsieck, Kenneth H., Kobulnicky, Henry A., Williams, Ted B., O'Donoghue, Darragh, Smith, Michael P., Percival, Jeffrey W., Prime Focus Imaging Spectrograph for the Southern African Large Telescope: optical design, 2003 SPIE, 4841, 1463. [2] Stobie, Robert, Meiring, Jacobus G., Buckley, David A. H., Design of the Southern African Large Telescope (SALT), 2000, SPIE, 4003, 355. [3] Sheinis, Andrew I., Wolf, Marsha J., Bershady, Matthew A., Buckley, David A. H., Nordsieck, Kenneth H., Williams, Ted B., The NIR upgrade to the SALT Robert Stobie Spectrograph, 2006, SPIE, 6269, 62694T. [4] Wolf, Marsha J., Sheinis, Andrew I., Mulligan, Mark P., Wong, Jeffrey P., Rogers, Allen, Designing the optimal semi-warm NIR spectrograph for SALT via detailed thermal analysis, 2008, SPIE, 7014, 701432. [5] Smith, Michael P., Wong, Jeffrey P., Mason, William P., Adler, Douglas P., Rogers, Allen R., Sheinis, Andrew I., Wolf, Marsha J., Thielman, Donald, J., Mulligan, Mark P., Percival, Jeffrey W., Mechanical design of the near-infrared arm of the Robert Stobie spectrograph for SALT, 2010, SPIE, 7735 (paper 7735-274 this conference). [6] RCWA codes developed by Gary Bernstein, http://www.physics.upenn.edu/~garyb/. [7] Rangwala, Naseem, Williams, T.B., Pietraszewski, Chris, Joseph, Charles L., An imaging Fabry-Perot system for the Robert Stobie Spectrograph on the Southern African Large Telescope, 2008, Astronomical Journal, 135, 1825.